GPT-6 Astra: A new generation of intelligence

Benchmark Model Rank Results
abstract-reasoning-on-arc-agi-2GPT-6 Astra#1Accuracy: 95.0
agents-on-terminal-bench-science-0-1GPT-6 Astra#1Accuracy: 64.6
computer-use-agents-on-screenspot-proGPT-6 Astra (no tools)#1Accuracy: 92.7
exploit-generation-on-exploitbenchGPT-6 Astra#1Coverage: 1.0
multiple-choice-question-answering-mcqa-on-gpqa-diamondGPT-6 Astra#1Accuracy: 96.0
deep-research-agents-on-browsecompGPT-6 Astra#2Accuracy: 91.5
abstract-reasoning-on-arc-agiGPT-6 Astra#3Accuracy: 98.5
code-generation-on-terminal-bench-4-0GPT-6 Astra#3Accuracy: 57.9
medical-question-answering-on-healthbench-professionalGPT-6 Astra#3Score: 63.4
code-generation-on-deepsweGPT-6 Astra#5Pass@1: 0.741